Terrafuse AI, a climate and weather risk forecasting tool developer, has launched a new platform for assessing and mitigating the growing property risk from wildfires in California, called Wildfire AI.
Nearly 4.5 million U.S. homes were identified at high or extreme risk of wildfire, with more than 2 million in California alone.
Recent data shows the average number of major US wildfires (over 40,000 acres) per year has risen 30% over the last 15 years, contributing to wildfire-associated costs growing from $1 billion per year in the 1990s, to $17.1 billion in 2020.
“I experienced firsthand the devastating impacts of wildfires when, a few years ago, my entire neighborhood in California was evacuated and much of it burned down,” said Hunter Connell, co-founder and CEO of Terrafuse AI.
“We also know that without more accurate insights about property risk, homeowners, businesses, and governments are facing impossible odds to plan for, mitigate, and escape these catastrophes in time.”
“Innovation in climate risk assessment is needed now more than ever to keep up with the volatile nature of climate change,” added Bonnie Lei, Head of Global Strategic Partnerships for Microsoft AI for Earth. “Terrafuse AI is able to bring incredible speed and accuracy to climate risk analytics and chose to use Azure to accelerate the scaling of their solution.”
Terrafuse AI’s approach leverages machine learning models trained by more than 50 environmental conditions, including hyperlocal wind speed, vegetation characteristics, humidity and observed wildfire events.
These complex models produce an accurate measure, expressed as a risk score or annual exceedance probability, that quantifies the likelihood of a catastrophic wildfire event happening down to a scale of 100 feet.
Terrafuse AI’s models have been validated against $1 billion of proprietary insurance claims data.
“Unfortunately, 2021 has brought us yet another devastating fire season in California with elevated fire risk that results in fires exploding at numerous locations across the state at nearly the same time,” said Dr. Daniel Feldman, Atmospheric Scientist and Advisor to Terrafuse AI. “Fire risk extends far beyond the areas burned and is highly-localized and ever changing.”
“In the Caldor Fire, the Wildfire AI model’s prediction rapidly changed from low to a very high risk score in the span of just a few days during the rapid growth of that fire,” continued Feldman. “The spatial patterns of the risk score reflected the spatial patterns of the increased likelihood of fire due to the surface winds, fuels, and humidity conditions.”
